From: Proceedings of the AI and Manufacturing Research Planning Workshop. Copyright © 1996, AAAI (www.aaai.org). All rights reserved. High Performance Expert Systems Todd D. Cherkasky Workand TechnologyInstitute 1775 K St NWSuite 630 Washington, DC20006 tdc@wti.org Abstract This paper reviews two divergent approaches to the development and application of expert systems in manufacturingand suggests that AI researchers and system designerscan moreeffectively contribute to manufacturing solutions by pursuinghigh performancedesign rather than reinforcingtraditional technocentricassumptions. Introduction To make expert systems more relevant to the manufacturing environment, applied researchers and other artificial intelligence (AI) designers must face the actual problemsthat arise in manufacturing organizations. While current AI research responds to demandsfor specialization and flexibility by modeling and automating process planning, scheduling, material handling, and other complicatedprocesses, manufacturers’ attention has shifted from exclusively technical concerns to include the complex processes of workorganization, interpersonal interaction, and organizational learning. From design for manufacturing (DFM)to integrated process and product development(IPPD), organizational theorists, management scholars, and interdisciplinary design studies researchers recognize that manufacturing solutions require an understanding of the social, cultural, and institutional dynamics underlying technologies of production (Appelbaum & Batt 1994, Cormier 1994; Susman 1992). Al-based systems would better support manufacturing applications if they complemented manufacturers’ attention to such factors as work organization and teams and if they built systems that leverage and extend workers’ skills. Technocentrism The technocentric approach to manufacturing problems is a legacy of an industrial system that achieved production gains by increasing efficiencies in individual tasks, by concentrating information and control toward the top of a multi-layered hierarchy, and by adjusting humanactivity to the capabilities of new technologies (Howard& Schneider 22 M & Mimuf~cturing Workshop 1994). In technocentric development, investments in technology are assumed to compensate for inefficient, error-prone, and otherwise limited workers. Design proceeds linearly from conception to execution, from engineering to production. Those most closely affiliated with the design and development of machinery--engineers and managers--make all decisions about work organization, plant layout, and workers’ activity. Ignoring both the implications of increasingly competitive, segmentedmarkets and the affordances of new technology, technocentric designers overestimate what computers and machinerycan achieve on their ownand underestimate the social and organizational contributions to industrial production. High performance Over the last few decades, managementscholars have addressed some of the deficiencies of the tecimocentric model. In Technology and the Future of Work, Paul S. Adler highlights several key factors of successful production that are neglected by the technocentric approach: The effective use of newtechnologies will require a workforcewith both higher skills and broader roles .... (and) also requires that the business firm reconfigured to support a process of continuous learning .... It is becomingincreasingly obviousthat, to compete effectively, firms need to develop comprehensivelong-term strategies for building their "knowledgeassets," that is, for jointly managingthe interdependent development of the knowledge embodied in technical systems and the knowledge accumulated by employees. Clearly, the success of firms in developing this new approach depends to a great extent on changesin the broadersocial, political, and institutional context (1992,13). Situated in a larger arena than the narrowly focused technocentric model, new approaches to work systems--often characterized as high performance---emphasize continuous learning, a long-term orientation, and an From: Proceedings AI and Manufacturing Planning Workshop. Copyrightapplication © 1996, AAAI (www.aaai.org). reserved. increased relianceof the on workers’ discretionResearch and judgment. performance depends in part All onrights decisions made With a goal of improving overall organizational by applied researchers, designers, marketers, distributors, system integrators, and application engineers. (Individual effectiveness over the long term, high performance work systems foster workerparticipation in technologyselection, discretion is infiuenced to a great extent by institutional determinants as discussed below in "Entry Points for implementation and customization and provide training to improvetechnical and process skills. See Table I for the Change.") Thesedevelopers participate in either the design eight key elements of high performance, as enumerated by or implementation (or both) of expert systems economist Ray Marshall (1991). manufacturing environments. The design stage involves knowledge-acquisition and the merger of the knowledge domainwith an inference engine. Implementingthe expert systemrequires that it be integrated into the processes of Table 1 workalong with other manufacturing technologies. Eight Key Elements of High Performance WorkSystems !. Effective use of all companyresources, especially the insights and experienceof front-line workers,in order to Expert system design and implementation achievecontinuousimprovements in productivity. A dominantpart of the design of expert systems entails 2. Acuteconcernfor the quality of products and services in translating an expert’s knowledgeand experience into a order to satisfy the demands of a consumer-driven form that the expert system can digest. Knowledge marketplace. engineers have recognized this process as a serious Aparticipative and non-authoritarianmanagement style in . bottleneck that slows expert system development. which workers ... are empoweredto make significant Anthropologists that have studied the knowledge decisionsby (a) using their individualdiscretion,experience acquisition process suggest "that the knowledgeengineers’ and creativity, and (b) cooperatingwith their peers in assumptions have some unintended negative consequences mutuallysupportiveatmosphere. for their ownpractice, for the systemsthey build, and thus 4. Internalandexternalflexibility in orderto: (a) rapidlyadjust internal productionprocessesto producea variety of goods (potentially at least) for the broader society" (Forsythe or services; and (b) accurately comprehend the external 1993, 448). These are assumptions about the nature of environmentand adjust to changingeconomicand social knowledge, the capabilities of expert systems, and the trends. contributions that workers can make to manufacturing A positive incentive structure that includes: employment 5. applications. security; rewardsfor effectively workingin groups;decent Knowledge, as it is often viewed by knowledgepay and workingconditions; and policies that promotean acquisition engineers, is asocial and acultural---and thus appreciationfor howthe company functions as an integrated independent of human interaction. Essentially a whole. commodity, knowledge is elicited from a domain expert by 6. Leading-edgetechnologydeployedin a mannerthat extends knowledge engineers in what they consider to be a humancapabilities and builds uponthe skills, knowledge, straightforward, common-sense question-and-answer andinsights of personnelat all levels of the company. procedure. Knowledge, in this sense, is captured and 7. A well-trained and well-educated workforcecapable of: stored in an ahistorical context for later retrieval and improvinga company’swork organization and production application to manufacturing problems that have well processes; adapting existing machinetechnology and selecting new equipment; developing new and improved defined parameters. Because "knowledge engineers tend products and services; and engaging in continuous to reify knowledge, often leading to naive and learning... inappropriate elicitation strategies in dealing with experts," Anindependentsourceof powerfor workers---alabor union 8. they contribute to their ownfi’ustrations (Forsythe and collective bargaining agreement--that protects Buchanan 1992, 206). employeeinterests in the workplace; helps to equalize Beyondthe practical bottleneck that is created, more power relations with management; and provides severe implications for manufacturing systems result when mechanisms to resolve disagreementsthat arise becauseof expert systems are designed without recognition of their the inherently adversarial nature of labor management limitations. Onelimitation is an over-reliance on formal relations. models as the exclusive means for solving design problems. Although these models work particularly well Expert System Development for clearly defined problemsthat anticipate all potential contingencies, the complexity of the contemporary The contrast between technocentrism and high manufacturing environment demands workers’ skills and performancecan be seen in the developmentof a variety of innovations. technologies, including expert systems. Whether the enduser expert systemis configured as a technocentric or high Cherkasky Z3 Workers have compensatedfor inelegant designs and limits in engineering knowledge by employing a range of skills and experience-based knowledge, or even by unsophisticated intervention .... Althoughthe performanceof one or several machinesmight be well understood, when systems are truly integrated and various machinery and electronics interact directly with each other, a higher order of complexity is created. One must predict not only the function and limits of one machinebut the conditions created from the interaction of multiple machines and control systems; sometimes effects are created that arise solely from their interaction and could not be found in the operation of one unit in isolation. (Salzman Lurid 1995, 331) of production before someonerealized that something was wrong.... It was just too easy to trust that the computers had got it right.’ In addition, the opaque computerized systems prevented workers from learning howthey might improveoperations." (Upton 1995, 80) As operators’ direct experience with manufacturing processes declines, the long-term viability of knowledge domains deteriorates. The danger is that knowledge in expert systems adversely affects manufacturing applications because it remains static (as contrasted with knowledgethat is modified by social negotiation), brittle (it neglects common-senseknowledgeabout the world and contrasting views of different experts), and narrow (when KBSknowledge does not cover a broad enough domain because only one domainexpert is used in the elicitation process). (Forsythe 1993) If, in contrast, knowledge engineers understand the value of workers’ experience, they will design systems that dependupon and help augmenttheir skills, thereby making the knowledgebase more robust. Whenthese systems are linked to the rest of the manufacturingapplication, they provide the meansfor workersto contribute their ideas and to be actively engagedwith production processes, thereby ensuring the knowledge-domain’slong term relevance and viability. in the design of expert systems, knowledgeengineers and applied researchers shape what counts as relevant knowledge. Similarly, when implementing expert systems, plant-floor supervisors, operators, managers and application engineers makecrucial choices that determine the extent to which expert systems conform to technocentric or high performance tenets. The resulting expert system is then linked with other production technologies (e.g., graphical user interfaces, intelligent sensors, supervisory process control software, data acquisition boards, etc.) to become part of the manufacturing environment. A popular vendor in the machine control industry, Intellution ©, has recently acquired Visual Expert ©, a MicrosoR Windows©-based development and run-time package. Combiningthe graphical control interface and data acquisition power of Intellution’s FIX DMACS with Visual Expert’s decision support, lntellution provides a flexible product. Visual Expert can be configured as specified in its product announcement:"It provides stepby-step graphical representations of diagnostic procedures for equipment troubleshooting and minimises (sic) the possibility of humanerror by requiring the operator to log in and follow the instructions displayed on screen" (Expert Systems1995). In this technocentric configuration, frontline workers execute procedures designed in advancewith no opportunity to modify them. Visual Expert’s "display builder" and development environment--where the logic structure of pertinent processes are defined---are From: Proceedings of the AI and Manufacturing Research Planning Workshop. Copyright © 1996, AAAI (www.aaai.org). All rights reserved. Workers’ contributions to production are necessarily overlooked if the knowledge base assumes they do not exist. Their skills and experience will be neglected if knowledge engineers do not realize that knowledge is "distributed through the division of labor, the procedures for getting things done, etc." and that "complete and unambiguous knowledge about expert procedures is unlikely to be transmitted through experts’ verbal or written self-reports." (Forsythe 1993, 453) Defining the role of the eventual operator is thus a central issue for designers of expert systems. Accordingto Jfirgen Dorn of the Christian Doppler Laboratory for Expert Systems, the steel "industry has always been very innovative in the application of new production technologies and the latest computertechnology.... Despite this rapid automation, the process operator has remainedan important link in the production process, and with the introduction of these new control systems, operators are being overloaded with process data." (Dora 1996) The operator, increasingly removed from direct experience with the production process, faces a wealth of information generated by programmable controllers, sensors, and other intelligent 1/O devices. To deal with this data bottleneck, Dornsuggests that the industry should use "an expert system that acts as a consultant or even as a decision maker .... " Here is the crucial choice. Will systems be developed to leverage or to replace workers’ experience? While a high performance orientation to manufacturing actively encourages workers’ critical input and experience, the teclmocentric path consistently undervalues humancontributions to production. Emphasizingthe dangers of the technocentric approach, David M. Upton summarizes in the Harvard Business Review the case of MeadCorporation’s Escanaba paper mill: "The mill’s managers ... realized that computerization in itself was no panacea. Someof the machines in the mill were relatively highly computer integrated. ’Even though we had a lot of computers on the plant (sic), sometimeswe’dmakea couple of hours’ worth 24 M & Manufacturing Workshop unavailable to operators. The Visual Expert application manufacture(DFM).Thesemanufacturingtrends focus From: Proceedings of the AI and Manufacturing Research Planning Workshop. Copyright © 1996, AAAI (www.aaai.org). All rights reserved. could be configureddifferently, allowingoperatorsto add organizational issues that traditional technocentric nodes andproceduresto its logic structure as well as to approaches to manufacturingignore. IPPDand DFM,for perform the more passive task of monitoring alarms, example, stress the importance of communicationand temperature, and other process information. The high teamsandtheygrant authorityto thosetraditionallyleft out performanceimplementationof Visual Expert allows users of design processes (e.g., production and maintenance to be actively engagedwith the process, to update the workers, as well as customers) (Thomas1994; Liker systemwith real time processimprovements, and to ensure Fleischer 1992). that ’the computers got it right.’ Theseinitiatives recognizethat continual innovations require the experiential knowledgeof humanexperts-expertise that is based on direct interaction with Technocentric design and manufacturing trends manufacturing processes. They draw on economicand Oftenexpert systemdevelopersassumethat manufacturing intrinsic rewardsthat stemfromparticipatory andskillapplications are moreeffective if they are controlled based applications of technology. For example,a $500 entirely by "decision-malting"expert systemsrather than million plant that producesspecial order alloy steel was by workerswhoare assisted by them. This assumptionis explicitly designedto incorporatethe real-timeinteraction readily apparent in documentation summarizing the of the experiencedoperators. Thefundamentalassumption benefits providedby expert systems. For example,General Electric’s Dodger--adiagnostic expert system--"employs of this design is that the computer makes recommendations, but does not make decisions a numberof AI techniquesto analyzeeddycurrent signals automatically.It provideson-line diagnosticcapabilities andassess trends in material condition. Benefits realized andrelies onthe critical insightsof plantpersonnel. include a high consistency of interpretation and a decreased reliance on humanoperators". (KIahr & Bymes The design of the system was based on an explicit 1993) The implicit (and sometimesexplicit) statement principle of operator control of the process. The system was designed to provide operators with the being madeby these researchers and the technologiesthey informationand technologynecessaryto makenearly produceis that operators, skilled machinists, process all production decisions without supervision. It engineers,cost estimatorsand other workersandtechnical providedoperators with a display showingthe status professionals are liabilities to the productionprocess. of all the heats. At each operation, it provided a Thesepersonnelare seen as expensivevariable costs, as computer-calculated "recommended" procedure and inefficient anderror-prone, or as lacking the specialized an analysis of any operator-proposedchanges.There expertise of moreexperiencedemployees(Chorafas1992; werealso manualoverrides for all equipmentin the Cunningham& Smart 1993; Hayes-Roth & Jacobstein process.... (Salzman 1993,90) 1994;Hauser&Hebert1992;Kowalskiet al. 1993). Although eachof these ostensibleliabilities offers ample The benefits included "a flow of process improvement opportunity for technical problemsolving, the current ideas" and total system performance that exceeded manufacturingenvironmentdemandsan alternative to the productivity and quality expectations and industry technocentric strategy of design. Current manufacturing standards. Severalrecent studies concur:participation in trends recognizethat labor is not the only---andnot the workplacedecisions, including technology design and most significantmvariable cost. Other costs in this implementation,increases productivity (Cormier1994; category include inventory, purchase price, machine Potter & Ngan1996; Levine& Tyson1990; Keefe1995). utilization, scrap, etc. Cycle-timereductionsand other These insights can be extended to knowledgeprocess improvements stem fromcontinual correctives to engineering,resulting in high-performance expert systems workorganization and technologyconfiguration, which that serve the needs of experts, relying on and can be heavily influenced by the input of front-line supplementingtheir skills instead of replacing them. workers. In high performance manufacturing Knowledge in "anthropocentric production systems"--the environments,workersat all levels of the companyare Europeanequivalent of high performance---isunderstood understoodas innovative contributors to both design and to include muchmore manufacturing solutions. To ensure continual whichhas been left to waste and has not ... been contributions from these workers, high performancework exploited yet: creativity, imagination,flexibility, systemsemphasizetraining for both technical and process learning frommistakesandthe potential to copewith skills. (SeeTable1.) the unforeseenand deal with contingencies. Manyof The organization-centered approach that looks to these qualities cannot be programmed in advancenor workersfor process innovationsis realized in integrated elicited by tight proceduresof command and control product and process development(IPPD)and design for or assignedto strictly prescribedtasks nor conserved CherkaskT 25 Three ©key institutions that can helpreserved. the design in the software computerised (sic) work (Littek From: Proceedings of the of AI and Manufacturing Research Planning Workshop. Copyright 1996, AAAI (www.aaai.org). All rights Charles 1995, 6). Derivedfrom the results of newmanufacturingpractices, a high-performance, skill-based understanding of what knowledgeis considered relevant and howexpert systems are implementedwill ensure the long term viability of the knowledge domain. Entry points for change To ensure the design of high performance expert systems, which criteria must applied researchers, knowledge engineers, system integrators, and application engineers follow? Given the limitations of technocentric design and the performance advantages of skill-based, participative design, we might ask designers to: ¯ Focus on process and product quality ¯ Ensure long-term inputs to the system ¯ Utilize and augmentworkers’ skills Referring to Table 1, each of the eight key elements of high performance work systems relates to at least one of these three criteria. The concern for quality relies on a high performance approach, which emphasizes overall organizational performance through both process and product improvementsinstead of a bias towards reducing costs and increasing individual efficiencies (Table I, items 1, 2 and 5). To ensure long term inputs to the knowledge domain and end-user system by experienced and skilled users, designers must count elements of organizational culture and workers’ knowledgeof day-to-day operations as relevant to the knowledgedomain. Over the long term, workers need to he trained to perform at expert levels to enable continual innovations (Table 1, items 1, 3, 4, and 7). The high performance expert system utilizes and augments workers’ skills by providing diagnostic support and integrating humaninsights interactively with machine "intelligence" instead of pursuing automatic control with operators only reacting to alarms in a crisis (Table !, items 1 and 6). Thesecriteria for high performancedesign are subject to value orientations and structural incentives beyond the level of individual designers. "The combination of a country’s or a firm’s economic market, employment practices, education system, legal system, social and engineering values, and other factors shape the wayit will organize its production process or, for our purposes, technology design." (Salzman & Rosenthal 1993, 46) shift in design criteria thus requires not only the participation of individual designers, but also organizational and institutional changes. 26 AI & Manufacturing Workshop environment to prioritize high performance criteria over technocentric ones are government,professional societies, and labor unions. Government’s role is to help guide technological development by providing public funds and institutions for basic and applied research, amongother resources and regulations. Ultimately, however, technologies do not develop autonomouslyguided by some natural internal logic; and thus governmentcan not merely prod technologies along a single linear trajectory. Althoughthere are significant constraints to social control over technological development’, choices madeby funding agents, researchers, and designers facilitate certain social arrangements and constrain others. Government, as a means for reconciling diverse values and opinions, must make normative assessments of what technologies should be developed and provide a research agenda that supports all of the society’s stakeholders. Withthis responsibility in mind, top-level policy makers from the federal governmenthave been emphasizing a high performance approach to technological innovation and new work systems. They encourage workers to master knowledge-intensive skills via continual training and education, as a way to help them maintain secure jobs in the global marketplace. In general, they see technical, specialized knowledgeas the key to personal, corporate, and national success. Secretary of LaborRobert Reich, for example, sees a promising future for "symbolic analysts," knowledge workers who leverage their value to the globally competitive firm based on problem-solvingskills and data analysis. (Reich 1991) In Technology for America’s EconomicGrowth, the Clinton administration recognizes that "the new growth industries are knowledge based. They depend on the continuous generation of new technological innovations and the rapid transformation of these innovations into commercial products the world wants to buy. That requires a talented and adaptive workforce capable of using the latest technologies and reaching ever-higher levels of productivity." (Clinton Gore 1993) These technologies, including knowledgebased systems, can be developed with either technocentric or high performance assumptions. One entry point for change towards the latter is a technology policy for high performance design. (Jarboe & Yudken1996) Another significant means for influencing the design context involves designers as professionals. Individually and through professional associations, designers are in a good position to contribute to the high performanceagenda 1e.g., technological momentum, inherent physical properties of materials and machines, emergent organizational and institutional imperatives, etc. (Clark et al. 1988; Hughes 1983; Winner1977; Winner 1995) outlined here----one workorganization the Workshop. From: Proceedings ofthat the AIincludes and Manufacturing Researchand Planning Copyright © 1996, the AAAIstructure (www.aaai.org). All rights reserved. Besides providing for processes to run critical judgment of users. By broadening the design smoothly, unions offer a variety of resources. Froma problemto include both technical and social processes, local to internationallevel, throughstrategic partners and consultants, unionssupportworkers’needsto be effective individual technicalprofessionalscan join other designers froma variety of disciplinesto expandtheir skills. Often, participants in the design process. They encourage professionalsocieties serve this networking function. They continual learning for workersand increased training and also encourage joint work between, for example, education in both process and technical skills. Unions engineering societies and other professional groups, provide conferences and seminars on workplace including those fromthe social sciences and education. technologiesand can tap the resources of consultants on Professionalassociationsalso often shapeeducationaland workplace change and technology choices and training curricula for their members,affecting both applications. eventual and incumbent designers’ assumptions. New Finally, labor unionsprovidea sourceof stability and programs,in interdisciplinaryengineeringdesignstudies--long-term commitmentthat is not always found among informedby the participation of engineeringprofessionals managersand other professionals. In manycorporations, andthe endorsement of their societies---addressthe issues managersandengineersare rotated fromplant to plant as of re-training, a~actingunder-representedstudents to the part of the overall businessstrategy. In contrast, workers field, reducing the effects of artificial disciplinary and their unionrepresentatives are morelikely to spend boundaries, challenging traditional curriculum and their entire career in one location. They possess an extensive knowledgeof the everyday contingencies and accreditationstandards,discussingsocially relevant issues, and expandingexclusive, linear def’mitions of design interactions. In addition, becausetheir ownfinancial processes. (Schumacher1995) Reinforcing an expanded futures are tied to the successof the plant, they are deeply view of design through educational change, civic committed to its success. professionalismencouragesa high performance orientation to design. Accordingto WilliamSullivan, author of Work High Performance Expert Systems and andIntegrity"the role of professionalexpertise... is being Manufacturing redefined away from providing discrete technical interventions.Instead, professionalsare givingcontinuing This paper has identified two viable meansfor creating attention to the institutional and cultural infrastructure expert systems that solve practical problems in neededto sustain a modem society." (1995, 237) manufacturingapplications. It has outlined the One of these institutions, effective labor unions, deficiencies of the technocentricapproachto expert system provides the necessary resources and structure to help development and highlightedthe benefits of pursuinghigh developersmeeteach of the three designcriteria for high performancedesign. performanceexpert systems:quality, long-terminputs to Robert Thomas,an MITSloan School of Management the knowledge domain and larger manufacturing scholar on manufacturingand organizational studies, application, and an emphasison workers’ skill. Unions confirmsthat "the humanand social systemsthat makeup provide an environmentwhere workers can feel secure organizations... can be eliminated. That is, workcan be enoughto critique existing procedures, to contribute to automated.Organizationalpolicies can be tightened and innovationsthat increase performance.(Table 1, item 8.) moreclosely monitored.People can be madeto go away. Withan explicit recognitionof the variousinterests in the But the social costs of unemployedand underemployed workplace, workers and designers will more seemlessly people and skills will be quite high ... and ... negotiateinevitable disagreements becausethey will havea organizational costs will take the form of a diminished reliable, well structured meanstowards solution. Exability to innovate--arisky strategy to pursuein a global Secretaryof LaborRayMarshallconcursthat if this laboreconomythat demandsinnovation and change." (Thomas managementdynamic "is good and effective, it is 1994, 247) inherently adversarial. In the best of circumstances,both Expert systemdevelopersmustrealize, as Dorahas for sides havedifferent perspectivesandare goingto disagree the steel industry, that "despite even more pervasive about somethings. Howhigh are mywagesgoing to be? automation in the future, humanexperts will remain Whose going to run things? Disagreement is inevitable and unavoidable for production control .... because new healthy. But a functioning adversarial relationship productionfailures that cannotbe handledadequatelyby a providesa wayto resolve differences."(Marshall1991,12) system occur quite regularly." (Dora 1996, 21) Equally important for the design context, the union production operations require the participation and provides a liaison between workers and engineers---a intervention of humanexperts, expert systemsshould be structured process for regular, two-waycommunication. Cherkasky 27 designed to operate interactively with workers and to complement,not replace, their knowledgeand experience. In current manufacturing applications, however, developers of expert systems too easily impart the technocentric perspective that humansare the weaklink. In doing so, they limit options for the design of systems using intelligent machines--resulting not only in harm to workers, but also to their firms, in 1988 ShoshanaZuboff published in the Age of the Smart Machine,a classic study that highlighted how workers could take advantage of computer-based control technology to contribute productively to the manufacturing process. Zuboff’s contribution to the study of work-life quality was to recognize the "informating" capabilities of computer-based programmablelogic controllers. She noted correctly that shop-floor technologies of the 1980s, while intensively automating manufacturing processes, actually generated information about the processes they controlled. So, Zuboff suggested that instead of de-skilling workers, managers ought to encourage workers to develop intellective skills--the ability to reason and analyze information. Although the capabilities of new manufacturing technologies continue to increase--even intellective skills can now be handled by intelligent machines--their implementation is not determined by the force of autonomous technological development. These technologies are consciously developed, with decisions being made throughout the complex product development process. Therefore, Zuboff’s lesson for managers and designers is relevant today as well. She outlined the tendency for managers to design programmable logic control technology to reinforce technocentric assumptions. The managers, therefore, sacrificed opportunities for alternative arrangements that were more flexible, productive and able to synthesize the strengths of both humans andmachines. Since that time manufacturing managers have begun to recognize the value of high performance skill-based design. In contrast, the technocentric emphasis of current expert systems development suggests that humanneeds and contributions to the production process may be neglected, threatening future manufacturingproductivity and flexibility. Chorafas, D. N. 1992. Expert systems in manufacturing. NewYork: Van Nostrand Reinhold Publishers. From: Proceedings of the AI and Manufacturing Research Planning Workshop. Copyright © 1996, AAAI (www.aaai.org). All rights reserved. References Adler, P. S., ed. 1992. Technologyand the future of work. NewYork: Oxford University Press. Appelbaum, E. and Batt, R. 1994. The new American workplace: Transforming work systems in the United States. Ithaca, NewYork: ILRPress. 28 AI & Manufacturing Workshop Clark, J., McLoughlin,I., Rose, H. and King, R. 1988. The process of technological change: New technology and social choice in the workplace. Cambridge: Cambridge University Press. Clinton, B. and Gore, A. 1993. Technology for America’s economic growth: A new direction to build economic strength. Washington, D.C.: U.S. GovernmentPrinting Office. Cormier, D. 1994. Commentson the relationship between worker participation and productivity. Paper given at University of Notre Dame Department of Economics Institutions Workshop,October 5. Notre Dame,IN. Cunningham,A. and Smart, R. ! 993. Computer-aidedparts estimation. A I Magazine14(3):39-49. Dom,J. 1996. Expert Systems in the Steel Industry. IEEE ExpertI I(1): 18-21. Expert Systems. 1995. Added expert systems. Expert Systems 12(1):68. Forsythe, D. E. 1993. Engineering knowledge: The construction of knowledgein artificial intelligence. Social Studies of Science 23(3):445-477. Forsythe, D. E. and Buchanan, B. G. 1992. Nontechnical problems in knowledge engineering: Implications for project management. Expert Systems with Applications 5:203-212. Hauser, R. D. and Hebert, F. J. 1992. Managerialissues in expert system implementation. SAM Advanced Management Journal 57(1): I 0-15. Hayes-Roth, F. and Jacobstein, N. 1994. The state of knowledge-based systems. Communications of the ACM 37(3):27-39. Howard,R. and Schneider, L. 1994. Workerparticipation in technological change:interests, influence, and scope, in Critical studies in organization and bureaucracy,ed. by F. Fischer and C. Sirianni, 519-544. Philadelphia: Temple University Press. Hughes, T., Park. 1983. Networks of power. Baltimore, MD:Johns Hopkins. Jarboe, K. P. and of Yudken, J. Manufacturing 1996. Smart Research workers, Planning smart Workshop. Salzman, H. and S. R. 1993. Sotware design: From: Proceedings the AI and Copyright © Rosenthal, 1996, AAAI (www.aaai.org). All rightsbyreserved. machines: A technology policy for the 21st century, Shaping technology and the workplace. NewYork: Oxford Washington, D.C.: Work& TechnologyInstitute. University Press. Keefe, J. 1995. The role of transformational union leadership in workplace restructuring, BriefmgPaper, 9501, Washington, D.C.: AFL-CIO Human Resources DevelopmentInstitute. Schumacher,J. A. 1995. Engineering and society: The art of design, Interim Report, EW-20263-94, Troy, NY: RensselaerPolytechnicInstitute. Klahr, P. and Byrnes, E. 1993. Innovative applications of artificial intelligence 5. MenloPark, CA:AAAIPress. Sullivan, W.M. 1995. Workand integrity: The crisis and promise of professionalism in America. New York: HarberBusiness. Kowalski, A., Bouchard, D., Allen, L., Latin, Y. and Vadas, O. 1993. Pitch expert: A problem-solving system for kraft mills. AI Magazine14(3):81-98. Susman, G. I., ed. 1992. Integrating design and manufacturing for competitive advantage. NewYork: OxfordUniversity Press. Levine, D. I. and Tyson, L. D. A. 1990. Participation, productivity, and the finn’s environment. In Paying for productivity, ed. by A. S. Blinder, 183-243. Washington DC: The Brookings Institution. Thomas,R~ J. 1994. Whatmachinescan’t do: Politics and technology in the industrial enterprise. Berkeley: University of California Press. Liker, J. K. and Fleischer, M. 1992. Organizational context barriers to DFM.In Integrating design and manufacturing for competitive advantage, ed. by G. I. Susman.NewYork: OxfordUniversity Press. Upton, D. M. 1995. Whatreally makesfactories flexible? HarvardBusiness Review. Winner, L. 1977. Autonomous technology. MA:MIT Press. Cambridge, Littek, W.and Charles, T., eds. 1995. The newdivision of labour: Emerging forms of work organization in internationalperspective. NewYork: Walter de Gruyter. Winner, L. 1995. The enduring dilemmas of autonomous technique. Bulletin of Science, Technology, and Society 15(2-3):67-72. Marshall, R. 1991. The eight key elements of high performance work systems. In Proceedings of High Performance Work and Learning Systems, 3-14. September 26-27. Washington, D.C.: AFL-CIO Human Resources DevelopmentInstitute. Zuboff, S. 1988. In the age of the smart machine: The future of work and power. Basic Books. Potter, E. E. and Ngan, Y. K. 1996. Estimating the potential productivity and real wageeffects of employee involvement, Washington DC: Employment Policy Foundation. Reich, R. B. 1991. The work of nations. Vintage Books. NewYork: Salzman, H. 1993. Skill-based design: Productivity, learning, and organizational effectiveness. In Usability: Turning technologies into tools, ed. by P. S. Adler and T. A. Winograd, 66-95. NewYork: Oxford University Press. Salzman, H. and Lund, R. 1995. Engineering principles and production technology: Automation and skill-based design in the US. International Journal of Vehicle Design 16(4/5):314-338.